[v2,0/3] btrfs: enhancement to pass generic/563

Message

Qu Wenruo Feb. 12, 2025, 2:52 a.m. UTC

[CHANGELOG]
v2:
- Rebased to the latest for-next branch
  There is a data corruption read fix, which changed the timing of
  btrfs_lock_and_flush_ordered_range().

- Introduce a dedicated and smarted, read path speific extent range lock
  helper, lock_extents_for_read()
  Which has all the extra subpage specific deadlock avoiding mechanism,
  along with a much better comments on which type of ordered extents
  needs to be waited, and which can be completely skipped.

The test case generic/563 on aarch64 with 64K page size and 4K fs block
size will fail with btrfs, but not EXT4 nor XFS.

The detailed reason is explained in the last patch, the TL;DR is that
btrfs is not handling block aligned buffered write in an optimized way
for subpage cases (block size < page size).

The first patch is to address the deadlock-prone
btrfs_lock_and_flush_ordered_range() in read paths, by introduce a
deadlock-avoiding helper for read paths.

The second patch is a refactor in preparation for the new enhancement.

Eventually the last patch will enable the enhancement and pass the
generic/563.

Qu Wenruo (3):
  btrfs: introduce a read path dedicated extent lock helper
  btrfs: make btrfs_do_readpage() to do block-by-block read
  btrfs: allow buffered write to avoid full page read if it's block
    aligned

 fs/btrfs/defrag.c       |   2 +-
 fs/btrfs/direct-io.c    |   2 +-
 fs/btrfs/extent_io.c    | 224 +++++++++++++++++++++++++++++++++++-----
 fs/btrfs/file.c         |   9 +-
 fs/btrfs/inode.c        |   4 +-
 fs/btrfs/ordered-data.c |  29 ++++--
 fs/btrfs/ordered-data.h |   3 +-
 7 files changed, 229 insertions(+), 44 deletions(-)